4058
Biochemistry 2006, 45, 4058-4068
In Situ Single-Molecule Imaging with Attoliter Detection Using Objective Total Internal Reflection Confocal Microscopy† Thomas P. Burghardt,*,‡ Katalin Ajtai,‡ and Julian Borejdo§ Department of Physiology and Biomedical Imaging, Mayo Clinic, Rochester, Minnesota 55905, and Department of Molecular Biology and Immunology, UniVersity of North Texas HSC, Fort Worth, Texas 76107 ReceiVed October 14, 2005; ReVised Manuscript ReceiVed February 2, 2006
ABSTRACT: Confocal microscopy is widely used for acquiring high spatial resolution tissue sample images of interesting fluorescent molecules inside cells. The fluorescent molecules are often tagged proteins participating in a biological function. The high spatial resolution of confocal microscopy compared to wide field imaging comes from an ability to optically isolate and image exceedingly small volume elements made up of the lateral (focal plane) and depth dimensions. Confocal microscopy at the optical diffraction limit images volumes on the order of ∼0.5 femtoliter (10-15 L). Further resolution enhancement can be achieved with total internal reflection microscopy (TIRM). With TIRM, an exponentially decaying electromagnetic field (near-field) established on the surface of the sample defines a subdiffraction limit dimension that, when combined with conventional confocal microscopy, permits image formation from 75% of the bleachable fluorescence detected with fTIRM is emitted by five chromophore-labeled myosins in a muscle fiber.
Protein structure and function is widely studied using tools designed to isolate and characterize single molecules. The approach avoids certain ambiguities introduced by ensemble averaging, thereby providing substantially more insight into protein dynamics accompanying function (1, 2). Motor proteins operating on the substrate in vitro is an example of a system successfully investigated with the single-molecule approach (3, 4). While the approach is advantageous in many ways, single molecules in vitro do not fully exemplify the native system in a cell where molecular crowding is the norm (5, 6). A crowded environment potentially affects protein solubility, structure, and kinetics, suggesting that proteins in situ function differently from their in vitro counterpart. For instance, the crowded environment is known to produce preferential hydration of a target protein favoring lower surface area structures and promoting self-association. Singlemolecule studies conducted in situ are clearly preferable when feasible. † Supported by NIH-NIAMS Grant R01AR049277 and the Mayo Foundation (to T.P.B. and K.A.) and by R21CA9732 and R01AR048622 (NIAMS) (to J.B.). * To whom correspondence should be addressed. Telephone: 507284-8120. Fax: 507-284-9349. E-mail:
[email protected]. ‡ Mayo Clinic. § University of North Texas HSC.
An interesting target protein subjected to molecular crowding is the motor protein myosin in muscle. In the skeletal muscle sarcomere, myosin forms thick filaments interdigitated with polymerized actin thin filaments. The filaments produce contraction by a relative sliding movement accompanied by myosin hydrolysis of ATP and sarcomere shortening (7, 8). Thick filaments are concentrated into the sarcomeric A-band where the myosin concentration is estimated to be ∼120 µM (9), implying ∼72 myosins per attoL (10-18 L). Using fluorescence detection of an extrinsic chromophore specifically labeling ∼1% of the myosin molecules, ∼1.4 attoL observation volume is needed to isolate a fluorescent signal from a single sarcomeric myosin. AttoL volume resolution is beyond diffraction-limited optics, stimulating wide interest in alternative approaches to relax diffraction-limited constraints. The standard optical microscope employs a propagating field (or far-field) electromagnetic radiation for imaging and is a diffraction-limited instrument; however, there are some notable exceptions. Schrader et al. (10) demonstrated subdiffraction resolution using a 4Pi microscope with twophoton excitation. Coherent light, focused through two highaperture objectives subtending a large solid angle about the sample, constructively interferes at the focus to produce an intensity peak much narrower than that obtainable from a
10.1021/bi052097d CCC: $33.50 © 2006 American Chemical Society Published on Web 03/09/2006
In Situ Attoliter Detection in Objective TIR single objective but at the cost of additional side lobes displaced laterally from the focus. The confocal pinhole plus image processing effectively removes the side lobe contributions. They demonstrated a resolution on the order of 100 nm in three dimensions, corresponding to a detection volume probably somewhat larger than 1 attoL. Klar et al. (11) demonstrated a 0.67 attoL fluorescence detection volume in a far-field microscope using stimulated emission depletion (STED). Synchronized probing and STED laser beams excite fluorescence in a normal diffraction-limited volume and then quench fluorescence from the periphery of a volume using stimulated emission initiated by the STED beam that is redshifted from the probing beam. The intensity profile of the STED beam is controlled to engineer shape and size of the volume from which fluorescence is detected. A microscope using STED or a related depletion technique appears able to shrink the detected fluorescent volume size without limit. Recently, the 4Pi and STED technologies were combined to achieve an axial resolution of ∼50 nm (12). Near-field microscopes make use of a nonpropagating evanescent field that decays exponentially in space to define a subdiffraction-limited volume. Near-field scanning optical microscopy (NSOM) probes (2) and zero-mode waveguides (13, 14) squeeze light through a nanoscale aperture to produce an evanescent field confined to the immediate vicinity of the aperture. NSOM makes use of a tapered fiber optic tip scanned over the sample to produce subdiffractionlimited resolution images of the surface. Zero-mode waveguides are not scanned, but they define a tiny illuminated volume on the sample surface. Either near-field method produces detection volumes of ∼1 attoL Total internal reflection microscopy (TIRM)1 produces an evanescent field on the water side of the glass/water interface made by a coverslip contacting aqueous solution (15). Excitation is delivered through the glass side by propagating light incident on the interface at angles beyond the critical angle for total internal reflection (16). Prism and prismless forms of TIRM have been described (17). Prismless TIRM utilizes a high numerical aperture oil immersion objective to refract excitation laser light at supercritical angles for total internal reflection (18). The evanescent field decays exponentially in the direction normal to the interface. Usually, prismless TIRM is setup to illuminate a uniform field in object space (unfocused TIRM or uTIRM) using a laser beam focused at the back focal plane. Ruckstuhl and Seeger (19-21) developed prismless TIRM with a parabolic mirror objective (PMO) producing a focused spot on the sample in the aqueous medium (focused TIRM or fTIRM). They combined fTIRM with confocal microscopy and investigated dye diffusion characteristics with fluorescence correlation spectroscopy (22). The findings indicated that the detection volume produced with this setup is 0) triaxial ellipsoid centered on the origin with semi-axis lengths Lx × Ly × Lz giving effective Vd values of ∼3.4 and ∼6.5 attoL. Sensitivity of sphere diffusion characteristics to the shape and size of Vd was tested by computing the mean and rms deviation of residence times (eq 2) for a large variety of practical beam profiles. Table 1 summarizes the beam profiles considered and their Vd values. The diffractionlimited profile listed in Table 1 is a hypothetical construct made from combining the lateral profile from a diffractionlimited beam (when the BFP is fully illuminated as opposed to annular illumination, see Figure 1) with the axial profile of an evanescent field. Figure 4 shows the mean (9) and rms deviation (2) for the number and fluorescence-intensityresidence duration distributions for 20 nm sphere-simulated diffusion. Correlation coefficients, F, of 0.98 or 0.96 between the rms deviation or mean fluorescence-intensity-residence duration and Vd, compared to 0.80 or 0.52 for their counterparts from the number-residence duration distribution, show that fluorescence-intensity-residence duration distribution is the more reliable indicator of Vd size. The fluorescence residence duration distribution weighs heaviest events producing more fluorescence. Events produce more fluorescence when the single fluorescent sphere inside Vd traverses regions of higher excitation intensity closer to the TIR surface (at z ) 0 in Figure 3) and centered laterally inside the focused spot. These spheres will also tend to spend more time inside Vd; thus, it is expected that fluorescence intensity weighting produces a parameter more sensitive to the Vd size. The mean and rms deviations shown in Figure 4 are derived from simulated data without noise. When simulating real data, we include photon statistic effects (noise) contributed from the background and sphere fluorescence signals. On noisy data, we employ event detection (see EVent Detection) to distinguish noise fluctuations from sphere detection events, leading to a dramatic decline in the mean and rms deviation of sphere residence duration in Vd. This is expected because points in the time-dependent fluorescence trace where the sphere is resident in Vd but near the edges of the volume where it produces lower fluorescence are indistinguishable from points where the sphere lies outside Vd. Ambiguous points are counted as times when the sphere lies outside Vd. RESULTS The top panel in Figure 5 shows fluorescence photon counts from a 20 nm diameter fluorescent sphere diffusing through Vd ≈ 3 attoL for fTIRM illumination with a NA 1.45, 60× objective and 10 µm diameter confocal pinhole. Fluorescence collection sampling times were 5 µs. The bottom panel shows simulated data from a randomly walking
4064 Biochemistry, Vol. 45, No. 13, 2006
Burghardt et al.
FIGURE 3: Integrated point spread functions (IPSF) for fTIRM and uTIRM. A and B contain the IPSF for the intensity distribution in Figure 2 and a 10 µm pinhole as a function of the point source position (x, y, and z). A shows IPSF for the point source constrained in the focal plane (z ) 0). B shows IPSF for the point source constrained in an axial plane (x ) 0). The z-dimension dependence in B is exponential with a field depth of ∼100 nm. C and D contain the IPSF for uTIRM with a 5 µm pinhole and other conditions identical to that in A and B. Table 1: Computed Detection Volumes for Various Beam Profiles Vd (attoL)
conditions
3.1 3.4 6.0 6.5 15.2 18.0
fTIRM, 100 nm evanescent field depth, 5 µm pinhole fTIRM, 100 nm evanescent field depth, 10 µm pinhole fTIRM, 100 nm evanescent field depth, 25 µm pinhole uTIRM, 100 nm evanescent field depth, 5 µm pinhole fTIRM, 100 nm evanescent field depth, 50 µm pinhole diffraction-limited profile, 100 nm evanescent field depth, 50 µm pinhole (hypothetical) uTIRM, 100 nm evanescent field depth, 50 µm pinhole uTIRM, 210 nm evanescent field depth, 50 µm pinhole
61.5 104.4
sphere. Both panels show fluorescence time traces with 1 event initiated when photon counts exceed 2 and subsequently terminated when photon counts are < 1 (see EVent Detection in the Materials and Methods). The time per event is 10 µs in the top panel and 5 µs in the bottom panel. Time increases from left to right in Figure 5. If simulated data (bottom panel) were run backward in time, the trace would correspond to a 10 µs event. The time asymmetry is a consequence of the event detection bias that registers new events only when the detected intensity is too large to be from a noise source characterized by the average background light level. Measured and simulated data from time traces longer than those depicted in Figure 5 often had the sphere entering and leaving Vd several times while lingering in the vicinity (data not shown). The number and fluorescence residence duration distributions, Rn and Rf, were tabulated from measured fluorescence intensity time traces of 20 and 100 nm diameter fluorescent spheres diffusing in water. Illumination mode (uTIRM or fTIRM), pinhole diameter, and evanescent field depth were varied. We computed the residence time average and variance by eq 2. Results are summarized in Table 2. Table 3 shows simulated quantities for a comparison to those in Table 2. We find that measured sphere diffusion characteristics for fTIRM with the 10 µm pinhole are fully accounted for by the simulation, consistent with the 3.4 attoL size
FIGURE 4: Mean (9) and rms deviation (2) for the number- and fluorescence-intensity-residence duration distributions for 20 nm sphere-simulated diffusion. Linear regression correlation coefficients, F, indicate how closely two variables approximate a linear relationship to each other. Mean and rms deviation of the fluorescence-intensity-residence duration distribution is the more reliable indicator of Vd size.
estimate for Vd in Table 1. Measurement and simulation comparison for fTIRM with the 50 µm pinhole suggests that the actual Vd is somewhat smaller than the 15.2 attoL estimate in Table 1. We did uTIRM experiments on 20 and 100 nm spheres for two evanescent field depths by using coverslips at the TIR interface with different refractive indices. The maximal field depth corresponded to a quartz coverslip with a refractive index of 1.46. The minimal field depth corresponded to the standard crown glass coverslip with a refractive index of 1.5. The excitation light incidence angle on the TIR
In Situ Attoliter Detection in Objective TIR
Biochemistry, Vol. 45, No. 13, 2006 4065
FIGURE 5: Fluorescence photon counts per 5 µs sampling time interval from a 20 nm diameter fluorescent sphere diffusing through a 3.4 attoL detection volume. Illumination is from fTIRM with a NA 1.45, 60× objective and 10 µm diameter confocal pinhole. A shows experimental data, and B shows the simulation based on the random walk diffusion model. Both A and B show fluorescence time traces with 1 event initiated when photon counts exceed 2 and subsequently terminated when photon counts are
